HelloTech Energy Builds AI Customer Service Knowledge Base on Amazon Bedrock, Empowering Generative-AI-Assisted Support

Title

HelloTech Energy Builds AI Customer Service Knowledge Base on Amazon Bedrock, Empowering Generative-AI-Assisted Support

Content

Shenzhen-based portable-energy manufacturer 华宝新能 (HelloTech Energy) extended its customer-service platform with an Amazon Bedrock knowledge base and Claude 3 reasoning to handle overseas user enquiries across chat, SMS, email, social media and in-app channels. The AWS case study quotes neither HelloTech executives nor implementation partners by name. The case study attributes the architectural decision to the HelloTech customer-service team, which identified two concrete limitations in the prior support system and worked with 亚马逊云科技 to design an extension layer. HelloTech's existing customer-support system aggregated overseas enquiries from five channels, but the system did not support custom user-intent labels (needed for operational analytics) and could only draw on knowledge provided by the vendor's ecosystem. To answer questions that required company-specific content, agents had to maintain a separate help-centre and route queries through manual lookup. HelloTech extended the existing system with a Bedrock-based plug-in. Customer-service teams can now drop knowledge documents into the Bedrock knowledge base and immediately have the heavy lifting—document chunking, embedding calculation and vector storage—handled automatically. The Bedrock knowledge base provides Retrieval-Augmented Generation (RAG) that automatically extracts, embeds and indexes company documents. A plug-in dispatches each support ticket to Claude 3 via Bedrock, matches the ticket against a predefined intent catalogue, classifies it as pre-sale, post-sale or repurchase, and applies a custom-label taxonomy. The knowledge base integrates with Bedrock Guardrails for input/output safety filtering and supports multi-turn conversational context management. Agents no longer need to maintain a separate help-centre or write manual intent rules for new ticket categories. The Bedrock knowledge base allows agents to ask natural-language questions against a single document and have the system synthesise answers, with lightweight human refinement before sending to overseas customers. The case study positions Bedrock's out-of-the-box RAG, multi-turn conversational context, Guardrails safety filtering and content moderation as the foundation for further extensions to HelloTech's overseas support workflows.